Laminar flow inhibits TNF-induced ASK1activation by preventing dissociation of ASK1from its inhibitor 14-3-3

Yingmei Liu, … , Bradford C. Berk, Wang Min

J Clin Invest. 2001;107(7):917-923. https://doi.org/10.1172/JCI11947.

The inflammatory cytokine TNF-a stimulates several presumed pro-atherogenic signalingevents in endothelial cells (ECs), including activation of c-Jun NH2-terminal kinase (JNK)and induction of E-selectin. Here, we show that apoptosis signal-regulating kinase 1(ASK1), a MAP kinase kinase kinase, is required for TNF-mediated JNK activation. TNFactivates ASK1 in part by dissociating ASK1 from its inhibitor 14-3-3. Because the risk ofatherosclerosis is decreased in regions of steady laminar flow, we hypothesized thatlaminar flow inhibits proinflammatory cytokine-mediated activation of JNK. Steady laminarflow inhibited both TNF activation of ASK1 and JNK. Inhibition of ASK1 by flow correlatedwith increased association of ASK1 with 14-3-3. A constitutively active form of ASK1 lackingthe 14-3-3-binding site (ASK1-ΔNS967A) was not inhibited by flow. These data establishASK1 as a target for flow-mediated inhibition of cytokine signaling and indicate a novel rolefor 14-3-3 as an anti-inflammatory mediator in ECs.

Article

Find the latest version:

http://jci.me/11947-pdf

IntroductionThe c-Jun NH2-terminal kinase (JNK) is one of threemitogen-activated protein kinase (MAPK) cascadesthat are strongly activated by stress signals and proin-flammatory cytokines such as TNF and IL-1. The TNFsuperfamily uses receptors that are devoid of intrinsiccatalytic activity (1). Activation of JNK by members ofthe TNF-receptor family is thought to be mediated bya family of intracellular signaling molecules known asTNFR-associated factors (TRAFs) (2). Six TRAF familymembers have been described, and different membersof this family display distinct receptor-binding speci-ficities. Studies from TRAF2 transgenic and knockoutmice demonstrated that TRAF2 is essential for activa-tion of JNK in response to TNF but not IL-1 (3–5). Incontrast, TRAF6 specifically mediates IL-1–inducedJNK activation by IL-1 receptor (6, 7).

JNK is activated by dual phosphorylation mediated byone of the MAP2Ks (MKK4 and MKK6). MKK, in turn,is activated through phosphorylation by MAP3Ks(including MEKK1, TAK1, and ASK1) (8–10). Some ofMAP3Ks such as MEKK1 and TAK1 can activate bothNF-κB and JNK cascades and these MAP3Ks are directtargets of TRAF molecules (6, 11, 12). However, someMAP3Ks such as ASK1, appear to be involved in JNKactivation only in response to proinflammatorycytokines and stress stimuli (13–15). ASK1 is a 170-kDaprotein that functionally is composed of an inhibitoryNH2-terminal domain, an internal kinase domain, anda COOH-terminal regulatory domain. The COOH-ter-minal domain of ASK1 binds to the TRAF domain, and

this association is required for ASK1 activation byTRAF2 and TRAF6 (15). A kinase-inactive form of ASK1(ASK1-K709R) functions as a dominant negative incytokine-induced JNK activation (13, 15). Deletion ofthe NH2-terminal 648 amino acids of ASK1 (ASK1-∆N)leads to constitutive ASK1 kinase activity as it does inother MAP3Ks, indicating that the NH2-terminus con-tains an inhibitory domain (16, 17).

In addition, several cellular factors, 14-3-3, andthioredoxin, have been reported to inhibit ASK1 activ-ity. Thioredoxin in a reduced form binds to the NH2-terminal part of ASK1 and blocks activation of ASK1by TNF (16–18); 14-3-3, a phosphoserine-binding mol-ecule, binds to ASK1 specifically via Ser-967 of ASK1and has been reported to inhibit ASK1-induced apop-tosis (19). However, the role of 14-3-3 in TNF-inducedASK1 activation has not been determined.

It is widely accepted that inflammation plays a key rolein the pathogenesis and progression of atherosclerosis(20). An important role for JNK in inflammation is sup-ported by many studies. First, JNK is activated by almostall proinflammatory mediators such as TNF, IL-1, LPS,and oxidative stress (10, 21). Second, JNK activation isessential for expression of many proinflammatory mol-ecules such as such as E-selectin, RANTES, IL-12, IL-6,and IL-8 by activating transcription factors including c-Jun and ATF-2 (10, 22–24). This has been demonstratedby either a dominant-negative or antisense approach.For example, overexpression of DN-JNK or antisenseJNK cDNA can block IL-6 and IL-8 expression inducedby IL-1 (25); overexpression of DN-JNK can block

The Journal of Clinical Investigation | April 2001 | Volume 107 | Number 7 917

Laminar flow inhibits TNF-induced ASK1 activation

by preventing dissociation of ASK1 from its inhibitor 14-3-3

Yingmei Liu, Guoyong Yin, James Surapisitchat, Bradford C. Berk, and Wang Min

Center for Cardiovascular Research, University of Rochester Medical Center, Rochester, New York, USA

Address correspondence to: Wang Min, Center for Cardiovascular Research, University of Rochester Medical Center, 601 Elmwood Avenue, Box 679, Rochester, New York 14642, USA. Phone: (716) 273-1499; Fax: (716) 275-9895; E-mail: wang_min@urmc.rochester.edu.

Yingmei Liu and Guoyong Yin contributed equally to this work.

Received for publication December 6, 2000, and accepted in revised form January 30, 2001.

The inflammatory cytokine TNF-α stimulates several presumed pro-atherogenic signaling events inendothelial cells (ECs), including activation of c-Jun NH2-terminal kinase (JNK) and induction of E-selectin. Here, we show that apoptosis signal-regulating kinase 1 (ASK1), a MAP kinase kinasekinase, is required for TNF-mediated JNK activation. TNF activates ASK1 in part by dissociatingASK1 from its inhibitor 14-3-3. Because the risk of atherosclerosis is decreased in regions of steadylaminar flow, we hypothesized that laminar flow inhibits proinflammatory cytokine-mediated acti-vation of JNK. Steady laminar flow inhibited both TNF activation of ASK1 and JNK. Inhibition ofASK1 by flow correlated with increased association of ASK1 with 14-3-3. A constitutively active formof ASK1 lacking the 14-3-3-binding site (ASK1-∆NS967A) was not inhibited by flow. These data estab-lish ASK1 as a target for flow-mediated inhibition of cytokine signaling and indicate a novel role for14-3-3 as an anti-inflammatory mediator in ECs.

J. Clin. Invest. 107:917–923 (2001).

endothelial cell (EC) surface expression in human umbil-ical vein endothelial cells (HUVECs) (23).

The concept that steady laminar blood flow exerts anatheroprotective effect by modulating EC function iswell supported (26, 27). Examples include increasednitric oxide production and decreased expression of celladhesion molecules such as VCAM-1 and E-selectin(26–28). In this study we demonstrate that activationof JNK, but not NF-κB, by proinflammatory cytokinesis attenuated by pre-exposure of EC to laminar flow. Togain insight into the mechanism by which laminar flowinhibits JNK activation by proinflammatory cytokines,we examined the effect of flow on the upstream activa-tors of JNK in TNF signaling pathway. Our resultsshow that flow inhibits ASK1 activity and ASK1-dependent JNK activation by increasing association ofASK1 with its inhibitor 14-3-3.

MethodsPlasmids. Mammalian expression plasmids for Flag-epi-tope labeled wild-type TRAF2 and dominant-negativeTRAF2 (dnTRAF2) (29) were provided by D.V. Goeddel(Tularik Inc., South San Francisco, California, USA);wild-type and the kinase-inactive ASK1 by GenhongChen (University of California, Los Angeles, California,USA); GST-14-3-3 by Anthony J. Muslin (WashingtonUniversity in St. Louis, St. Louis, Missouri, USA); GST-JNKK1 (MKK4) by Bing Su (M.D. Anderson, Texas,USA), pBIIXLUC plasmid (κB-LUC) (23), which con-tains two κB sites from the immunoglobulin kappaenhancer, and the JNK-dependent reporter gene system(FR-Luc and FA2-cJun) were from Stratagene (PathDe-tect Reporting Systems; La Jolla, California, USA).Expression plasmids for ASK1-∆N (lacking the NH2-terminal domain) and ASK1-∆NS967A (deletion of theNH2-terminal and mutation at serine 967) were con-structed into the Flag-vector.

Cells and cytokines. HUVECs are purchased from Clonet-ics Corp. (San Diego, California, USA). Human rTNF andrIL-1 are from R&D Systems (Minneapolis, Minnesota,USA) and used at 100 U/ml and 250 U/ml, respectively.

Flow experiments. Fluid shear stress in vitro was creat-ed by the cone and plate viscometer. Cells were grownon 60-mm dishes coated with gelatin. Upon reaching95% confluence, fresh media was added, and 2 dayslater cells were rinsed free of culture media with HBSS(containing, in mM: NaCl 130, KCl 5, CaCl2 1.5, MgCl2

1.0, HEPES 20, pH 7.4),with 10 mM glucose added, and10% serum, and either maintained in static conditionor exposed to flow (fluid shear stress = 12 dynes/cm2)in a parallel plate chamber at 37°C. After varying timesof exposure to flow, cells were washed gently with ice-cold PBS (composition, in mM: NaCl 137, KCl 2.7,Na2HPO4 4.3, KH2PO4 1.4, pH 7.3) and cell lysates werefor further analysis.

JNK and ASK1 kinase assays. JNK assay was performedas described previously (23) using GST–c-Jun (1-80)fusion protein as a substrate. ASK1 assay was per-formed using GST-MKK4 as a substrate.

Transfection and reporter assay. Transfection of HUVECswere performed by DEAE-dextran method, as describedpreviously (23). Luciferase activity followed by renillaactivity was measured twice in duplicate using aBerthold luminometer. All data were normalized as rel-ative luciferase light units/renilla unit.

Immunoprecipitation and immunoblotting. HUVECsafter various treatments were washed twice with coldPBS and lysed in 1.5 ml of cold lysis buffer (50 mMTris-HCl, pH 7.6, 150 mM NaCl, 0.1% Triton X-100,0.75% Brij 96, 1 mM sodium orthovanadate, 1 mMsodium fluoride, 1 mM sodium pyrophosphate, 10µg/ml aprotinin, 10 µg/ml leupeptin, 2 mM PMSF, 1mM EDTA) for 20 minutes on ice. For immunopre-cipitation to analyze protein interaction in vivo, super-natant of cell lysates were precleared by incubatingwith normal rabbit serum plus GammaBind plusSepharose beads on a rotator at 4°C overnight. Thelysates were then incubated with the first protein-spe-cific antiserum (e.g., 14-3-3; Santa Cruz Biotechnolo-gy Inc., Santa Cruz, California, USA) for 2 hours with50 µl of GammaBind plus Sepharose. Immune com-plexes were collected after each immunoprecipitationby centrifugation at 13,000 g for 10 minutes, followedby three to five washes with lysis buffer. The immunecomplexes were subjected to SDS-PAGE followed byimmunoblot analysis (Immobilon P; Millipore, Bed-ford, Massachusetts, USA) with the second protein-specific Ab (e.g., ASK1; Santa Cruz BiotechnologyInc.). The chemiluminescence was detected using anenhanced chemiluminescence (ECL) kit according tothe instructions of the manufacturer (Amersham LifeScience, Arlington Heights, Illinois, USA). For detec-tion of Flag-tagged proteins (ASK1-∆N and ASK1-∆NS967A), anti-Flag M2 Ab (Sigma, St. Louis, Mis-souri, USA) was used for immunoblot analysis. Fordetection of HA-tagged proteins (wild-type ASK1),anti-HA Ab (Roche Diagnostics, Indianapolis, Indiana,USA) was used for immunoblot analysis.

Electrophoretic mobility shift assays. The double-strandedoligonucleotide containing a κB consensus site fromthe immunoglobulin κ gene (Promega Corp., Madison,Wisconsin, USA) was used for electrophoretic mobilityshift assays (EMSAs). Preparation of nuclear extractsand EMSA were performed as described previously (23).

ResultsFlow inhibits JNK and ASK1 activation induced by cytokines.To examine the effect of flow on cytokine-induced JNKactivation, HUVECs were either maintained in staticconditions for 10 minutes or subjected to flow (shearstress = 12 dynes/cm2) for 10 minutes before TNF-α (100U/ml) stimulation for 15 minutes. TNF-induced JNKactivation was determined by an in vitro kinase assayusing GST–c-Jun as a substrate, as described previously(23). The results show that pre-exposure to flow signifi-cantly attenuated TNF-induced JNK activation. JNKactivity was also measured by determining phosphory-lation of endogenous c-Jun using Western blot analysis

918 The Journal of Clinical Investigation | April 2001 | Volume 107 | Number 7

with an Ab against phosphorylated c-Jun (23). A similareffect of flow on TNF-induced JNK activation wasobserved by this assay. In contrast to our findings, twoother laboratories (30, 31) observed JNK activation byflow. “Trivial” explanations for the differences include:we used HUVECs rather than bovine aortic endothelialcells; our cells were several days postconfluent; and ourcells were growth-arrested by contact inhibition whilethe other laboratories used serum deprivation.

To examine whether flow specifically inhibits JNK acti-vation by TNF, we determined the effect of flow on NF-κBactivation, a signaling pathway activated in parallel toJNK. We have shown previously that TNF activates main-ly the p65/p50 NF-κB complex in HUVECs (23). NF-κBactivation was measured by an EMSA using HUVECnuclear extracts. As shown in Figure 1b, pre-exposingHUVECs to flow did not inhibit TNF-stimulated NF-κBactivation. These data indicate that flow specificallyinhibits JNK activation by proinflammatory cytokines.

To examine the effect of flow on upstream activatorsof JNK, we measured the activity of ASK1, one of theMAP3Ks involved in activation of JNK, but not NF-κB.ASK1 activity was determined by an in vitro kinaseassay using GST-MKK4 (JNKK1) fusion protein as asubstrate. As shown in Figure 1c, TNF-α (100 U/mL)activated ASK1 maximally at 15 minutes in HUVECs.Pre-exposing HUVECs to flow for 10 minutes signifi-cantly inhibited TNF-stimulated ASK1 activity (80%inhibition; n = 3, P < 0.01).

ASK1 is specifically involved in JNK activation by TNF. Toexamine if ASK1 is a specific activator in JNK pathway inHUVECs, we performed JNK- and NF-κB–dependentreporter gene assays by transient transfection in ECs. Weused TRAF2 as a control, since we found previously thatTRAF2 is involved in TNF-induced activation of bothNF-κB and JNK in ECs (23, 32). Overexpression of dom-inant-negative TRAF2 (DN-TRAF2) blocked TNF-induced expression of JNK-dependent reporter genes

The Journal of Clinical Investigation | April 2001 | Volume 107 | Number 7 919

Figure 1Flow pre-exposure inhibits TNF-mediated JNK and ASK1, but not NF-κB, activation in ECs. ECs were subjected to the following “precon-ditioning” protocol: they were maintained in static conditions for 25minutes (Ctrl), exposed to flow for 10 minutes and then held staticfor 15 minutes (Flow), maintained in static conditions for 10 min-utes, followed by TNF stimulation for 15 minutes or subjected toflow for 10 minutes followed by TNF-α (100 U/ml) stimulation for15 minutes (flow + TNF). (a) Flow inhibits JNK activation by TNF.Cell lysates were prepared and analyzed for JNK activity by an in vitrokinase assay using GST–c-Jun as a substrate. (b) Flow has noinhibitory effect on NF-κB. Cells were treated as in a. Nuclear extractsfrom these cells were used for EMSA with a κB probe. Specificity ofNF-κB complex was verified by 50-fold molar excess of the unlabeledκB oligonucleotide in the TNF-treated sample (TNF + competitor).(c) Flow inhibits ASK1 activation by TNF. Cell lysates in a were pre-pared and analyzed for ASK1 activity by an in vitro kinase assay usingGST-MKK4 as a substrate. Autoradiograms shown in a–c are repre-sentative of three experiments in HUVECs.

Figure 2ASK1 is specifically involved in JNK activation by TNF. HUVECswere transiently transfected with the indicated expression con-structs (1 µg each) together with either JNK (a) or a NF-κκB (b) pro-moter-reporter gene (1 µg each). A constitutive expression vectorfor renilla unit (0.5 µg each) was also transfected for normalizationof transfection efficiency (see Methods). Cells were left untreatedor treated with TNF (100 U/ml TNF-α). Relative luciferase activi-ties (luciferase vs. renilla unit) from untreated or TNF-treated sam-ples are presented from the mean of duplicate samples. Similarresults were obtained from two additional experiments. The TNFresponses (fold induction) are shown. (a) Effect of kinase-inactiveASK1 (ASK1-K709R) or DN-TRAF2 on a JNK-dependent reportergene. (b) Effect of kinase-inactive ASK1 (ASK1-K709R) or DN-TRAF2 on a NF-κB–dependent reporter gene.

(Figure 2a) and NF-κB–dependent gene expression (Fig-ure 2b). In contrast, overexpression of kinase-inactiveASK1 dominant negative (ASK1-K709R) blocked bothTNF-induced expression of JNK-dependent reportergenes (Figure 2a), but not NF-κB–dependent geneexpression (Figure 2b). These data suggest that ASK1 isspecifically required for JNK activation by TNF.

TNF dissociates ASK1 from its inhibitor 14-3-3, whereas flowprevents dissociation of ASK1 from 14-3-3. We hypothesizedthat flow increases the interaction of ASK1 with itsinhibitor 14-3-3 to explain how ASK1 activation byTNF is inhibited by flow. Association of ASK1 with 14-3-3 was easily detected in untreated HUVECs (Fig-ure 3, control). TNF treatment significantly reducedthe interaction of ASK1 with 14-3-3, indicating thatTNF activates ASK1, in part, by dissociating ASK1 from14-3-3. In contrast, flow (using the preconditioningprotocol) prevented the TNF-induced dissociation ofASK1 from 14-3-3 (Figure 3).

ASK1-∆NS967A does not associate with 14-3-3. Whilethioredoxin binds to the NH2-terminal domain ofASK1 and 14-3-3 binds to the COOH-terminal ofASK1, both inhibit ASK1 activity. An NH2-terminalASK1 deletion mutant (ASK1-∆N) no longer binds tothioredoxin and shows increased kinase activity (16,17). Mutation of serine 967→alanine (ASK1-S967A)renders ASK1 defective in 14-3-3 binding and likelyincreases ASK1 activity as shown by accelerated ASK1-induced apoptosis (19). We constructed an ASK1“double” mutant, which lacks the NH2-terminaldomain and has a mutation of serine 967→alaninethat we termed ASK1-∆NS967A. To determinewhether ASK1-∆NS967A binds to these twoinhibitors, HA-tagged ASK1-WT, Flag-tagged ASK1-∆N and ASK1-∆NS967A were transiently transfectedinto HUVECs, and the interactions of these ASK1 pro-teins with thioredoxin and 14-3-3 were examined bycoimmunoprecipitation assay. As expected, ASK1-WTbound to both thioredoxin and 14-3-3 (Figure 4a).ASK1-∆N bound to 14-3-3 (lane 1 in Figure 4b), butnot to thioredoxin (lane 3 in Figure 4b). In contrast,ASK1-∆NS967A failed to bind either thioredoxin or14-3-3 (lanes 2 and 4 in Figure 4b).

ASK1-∆NS967A activity is not inhibited by flow. Todetermine the effect of altering ASK1 interaction with14-3-3, we measured both JNK activity and JNKreporter-gene expression. Expression of ASK1 pro-teins in transfected ECs was determined by Westernblot analysis with an Ab against the COOH-terminalASK1, which recognized ASK1-WT, ASK1-ASK1-K709R, ASK1-∆N, and ASK1-∆NS967A (Figure 5a).The endogenous ASK1 present in ECs was also detect-ed (see lanes 4 and 5). Expression of ASK1-WT, ASK1-∆N, or ASK1-∆NS967A increased JNK activity in theabsence of TNF as measured by an in vitro kinaseassay with GST–c-Jun substrate (Figure 5b) and JNK-dependent reporter-gene expression (Figure 5c).Expression of the kinase-inactive form of ASK1(ASK1-K709R) had no effect on JNK reporter-geneexpression (Figure 5c). These data confirm the func-tion of ASK1 protein in ECs.

Since ASK1-∆NS967A does not associate with 14-3-3,we hypothesized that ASK1-∆NS967A activity would notbe inhibited by flow. To test this hypothesis, ECs weretransfected with Flag-tagged ASK1-∆N or ASK1-∆NS967A constructs. Because ASK1-∆N and ASK1-∆NS967A are constitutively active, ECs were exposed toflow in the presence of TNF 24 hours after transfection.

920 The Journal of Clinical Investigation | April 2001 | Volume 107 | Number 7

Figure 3Flow enhances the interaction of ASK1 with its inhibitor 14-3-3.HUVECs were subjected to flow and TNF treatment as described inFigure 1. Cell lysates were immunoprecipitated (IP) with Ab against14-3-3 followed by an immunoblot (IB) analysis with anti-ASK1 (toppanel). The 14-3-3 protein in the immunoprecipitates was deter-mined by Western blot analysis with Ab against 14-3-3 (lower panel).

Figure 4ASK1-∆NS967A does not bind either thioredoxin or 14-3-3. HA-tagged ASK1-WT, Flag-tagged ASK1-∆N or ASK1-∆NS967A weretransiently transfected into HUVECs. Cell lysates were immunopre-cipitated by Ab’s as indicated (HA, 14-3-3, thioredoxin, or Flag).ASK1-WT was detected by Western blot analysis with anti-HA (a).ASK1-∆N and ASK1-∆NS967A were detected by anti-Flag (b).

Flag-tagged ASK1-∆N or ASK1-∆NS967A were immuno-precipitated by anti-Flag, and their activities were deter-mined by an in vitro kinase assay using GST-MKK4 as asubstrate. As shown in Figure 6a, ASK1-∆N expressionwas not altered by flow, but its activity, as measured byphosphorylation of GST-MKK4, was significantly inhib-ited by flow (Figure 6, b and c). In contrast, the activity ofASK1-∆NS967A was not inhibited by flow. These datasuggest that flow inhibits ASK1 activity by regulatingassociation of ASK1 with its inhibitor 14-3-3.

Flow and TNF modify phosphorylation of ASK-1 Ser-967.Ser-967 of ASK1 is part of a motif RxSxxP reported tobind 14-3-3; therefore, we hypothesized that TNF and

flow alter ASK1 phosphorylation at Ser-967 to regulatethe interaction of ASK1 with 14-3-3 (Figure 7a). To testthis hypothesis, HA-ASK1 transfected into ECs. ECswere exposed to flow or TNF 24 hours later. To meas-ure pSer-967 of ASK1 we developed an in vitro 14-3-3binding assay. Phosphorylated ASK1 Ser-967 was firstbound to GST-14-3-3, followed by Western blot analy-sis using anti-HA. This is different from the coim-munoprecipitation assay described in Figure 3, whichmeasured preexisting complexes of ASK1 and 14-3-3.We first examined function of GST–14-3-3 proteins.The results show that ASK1 specifically binds to wild-type 14-3-3, but not to a mutant 14-3-3 (K49A) defec-

The Journal of Clinical Investigation | April 2001 | Volume 107 | Number 7 921

Figure 5ASK1-∆N and ASK1-∆NS967A show increased JNK activation. (a)Overexpression of ASK1 proteins in ECs. HUVECs were transientlytransfected with control vector, ASK1-WT, ASK1-K709R, ASK1-∆N,and ASK1-∆NS967A, and cell lysates were used for protein expres-sion by Western blot with Ab against the COOH-terminal domain ofASK1. (b) Effect of ASK1 proteins on JNK activation. Cell extracts ina were used for in vitro kinase assay using GST–c-Jun as a substrate.(c) Effects of ASK1 proteins on JNK-dependent promoter. HUVECswere cotransfected with 1 µg of JNK reporter gene, a construct forrenilla unit (0.5 µg), and various ASK1 expression constructs (1 µgeach). The relative luciferase activity normalized to renilla unit isshown. Data are presented from mean of duplicate samples.

Figure 6ASK1-∆NS967A activity is not inhibited by flow. HUVECs were trans-fected with control vector, ASK-∆N, or ASK1-∆NS967A. Twenty-fourhours after transfection, cells were subjected to “preconditioning”protocol. Cell lysates were immunoprecipitated by anti-Flag. Theimmunoprecipitates were used to detect ASK1 by Western blot analy-sis with anti-Flag for protein expression (a) and to determine ASK1activity by an in vitro kinase assay using GST-MKK4 as a substrate(b). (c) The quantitative analysis of the radiogram in b. ASK1 activ-ity is shown as fold increase by taking vector transfection as one.Data are presented from mean of two independent experiments.

tive in phosphoserine-binding (33) (Figure 7b). Con-sistent with our observation in Figure 4, ASK1-∆N, butnot ASK1-∆NS967A, bound to GST–14-3-3 (data notshown). These data suggest GST–14-3-3 specificallybinds to phosphorylated ASK1 Ser-967 in an vitrobinding assay. Therefore, we use this GST–14-3-3 bind-ing assay to determine effects of TNF and flow onphosphorylation of ASK1 Ser-967. Our results showthat TNF decreased the amount of phosphorylatedASK1 leading to reduced binding on GST–14-3-3 (Fig-ure 7b). In contrast, flow prevents dephosphorylationof ASK1 at Ser-967, resulting in increased 14-3-3 bind-ing (Figure 7c). These data suggest that TNF and flowregulate ASK1 activity by modifying the phosphoryla-tion status of ASK1 Ser-967.

DiscussionThe major finding of the present study is that one of themechanisms by which steady laminar flow inhibitsproinflammatory events is via inhibition of ASK1.Based on findings in the present study, we propose amodel in which TNF activates ASK1 (in part) by disso-

ciating ASK1 from its inhibitor 14-3-3, while steadylaminar flow inhibits TNF-induced ASK1 and JNK acti-vation by preventing the release of ASK1 from 14-3-3(Figure 7). Because steady flow may exert atheroprotec-tive effects by inhibiting inflammation, these resultsimply an important role for 14-3-3 in atherosclerosis.

While some MAP3Ks such as TAK1 and MEKK1 areinvolved in both JNK and NF-κB activation, our dataindicate ASK1 is one of the major MAP3Ks specificallyassociated with JNK activation in ECs. It remains to bedetermined whether other MAP3Ks are regulated byTNF and flow in ECs.

The activation mechanisms for MAP3Ks in “thestress-activated” MAPK pathways remain largelyunknown. A role for reactive oxygen species (ROS) hasbeen proposed (17). Specifically, hydrogen peroxide acti-vates ASK1, and TNF-induced activation of ASK1 isinhibited by antioxidants (18). Thioredoxin, a redox-sensing protein, associates with ASK1 in its reducedform. Recently, it has been shown that TNF (as well asoverexpression of TRAF2) stimulates the production ofROS (16). Activation of ASK1 by TNF requires the ROS-mediated dissociation of thioredoxin followed by bind-ing of TRAF2 and subsequent ASK1 dimerization (16).The fact that flow also increases ROS in ECs (34, 35)suggests that thioredoxin is not a critical mediator offlow in regulating ASK1 activity. This is supported byour data that activity of the thioredoxin binding–defi-cient mutant (ASK1-∆N) is still inhibited by flow.

We found that 14-3-3 binds to ASK1 in unstimulatedHUVECs, indicating that ASK1 is constitutively phos-phorylated at Ser-967 and forms a preexisting complexwith 14-3-3 as described previously (19). A 14-3-3 mol-ecule with two mutations (R56A and R60A) is deficientin its ability to bind phosphoserine-containing peptidesand functions as a dominant-negative 14-3-3 signalingmolecule (33). Recently Xing et al. showed that expres-sion of DN-14-3-3 (R56A/R60A) activated ASK1 andJNK, likely by releasing ASK1 from inhibition byendogenous 14-3-3 (36). This is consistent with ourfinding that proinflammatory cytokines stimulaterelease of ASK1 from 14-3-3, whereas flow inhibits therelease. Currently the mechanism by which cytokinesand flow regulate association of ASK1 with 14-3-3 is notunderstood. However, our data suggest that phospho-rylation of ASK1 Ser-967 (which is the 14-3-3 bindingsite) is the critical step. We suggest that TNF activates aphosphatase that diminishes 14-3-3-ASK1 interaction,whereas flow inhibits this phosphatase to enhance for-mation of 14-3-3-ASK1 complex (Figure 7). Our findingthat activity of the 14-3-3 binding-deficient mutant(ASK1-∆NS967A) is not inhibited by flow further sup-ports this model. Experiments to identify this phos-phatase are underway.

In summary, these data establish the ASK1-JNK sig-naling cascade as a target for flow-mediated inhibitionof proinflammatory cytokine signaling and indicate anovel role for 14-3-3 as an anti-inflammatory mediator.While the roles of ASK1 and JNK in atherogenesis have

922 The Journal of Clinical Investigation | April 2001 | Volume 107 | Number 7

Figure 7Flow and TNF modify phosphorylation of ASK1 Ser-967. (a) A modelfor regulation of ASK1 activity by TNF and flow. ASK1 Ser-967 is ina phosphorylated state (pSer-967). TNF induces dephosphorylationof pSer-967 to release ASK1 from inhibitor 14-3-3. ASK1 dimeriza-tion results in activation of the ASK1-JNK cascade. Flow preventsTNF-induced dephosphorylation of pSer-967 to block release ofASK1 from 14-3-3. (b) GST–14-3-3 binding assay for phosphoryla-tion of ASK1 Ser-967. HA-tagged ASK1-WT was transfected intoECs, and pSer-967 of ASK1 was determined by an in vitro binding toGST–14-3-3 followed by Western blot analysis using anti-HA. GST,GST–14-3-3-K49A (a phosphoserine binding–deficient mutant) wereused as controls. (c) Flow and TNF regulate phosphorylation ofASK1 Ser-967. ECs were transfected with ASK-WT. Twenty-fourhours after transfection, cells were subjected to “preconditioning”protocol. Amount of phosphorylated ASK1 Ser-967 was measuredby an in vitro GST–14-3-3 binding assay. Data are representative oftwo experiments in HUVECs.

only been studied to a limited extent, this study sug-gests that inhibition of ASK1-JNK pathway may pro-vide a valid approach for antiatherosclerotic therapy.

AcknowledgmentsWe thank D.V. Goeddel for TRAF2 and DN-TARF2 con-structs; Genhong Chen for wild-type and the kinase-inactive ASK1; Anthony J. Muslin for GST-14-3-3;Stephanie Lehoux for GST-14-3-3 K49A; Bing Su forGST-JNKK1 (MKK4). We also thank B.C. Berk labmembers for assistance and discussions. This work wassupported by NIH grants 1R01HL-65978-01 to W. Minand 5P01HL-18645 to B.C. Berk.

1. Baker, S.J., and Reddy, E.P. 1998. Modulation of life and death by theTNF receptor superfamily. Oncogene. 17:3261–3270.

2. Arch, R.H., Gedrich, R.W., and Thompson, C.B. 1998. Tumor necrosisfactor receptor-associated factors (TRAFs): a family of adapter proteinsthat regulates life and death. Genes Dev. 12:2821–2830.

3. Lee, S.Y., et al. 1997. TRAF2 is essential for JNK but not NF-kappaBactivation and regulates lymphocyte proliferation and survival. Immu-nity. 7:703–713.

4. Lomaga, M.A., et al. 1999. TRAF6 deficiency results in osteopetrosisand defective interleukin-1, CD40, and LPS signaling. Genes Dev.13:1015–1024.

5. Yeh, W.C., et al. 1997. Early lethality, functional NF-kappaB activation,and increased sensitivity to TNF-induced cell death in TRAF2-defi-cient mice. Immunity. 7:715–725.

6. Baud, V., et al. 1999. Signaling by proinflammatory cytokines:oligomerization of TRAF2 and TRAF6 is sufficient for JNK and IKKactivation and target gene induction via an amino-terminal effectordomain. Genes Dev. 13:1297–1308.

7. Cao, Z., Xiong, J., Takeuchi, M., Kurama, T., and Goeddel, D.V. 1996.TRAF6 is a signal transducer for interleukin-1. Nature. 383:443–446.

8. Derijard, B., et al. 1994. JNK1: a protein kinase stimulated by UV lightand Ha-Ras that binds and phosphorylates the c-Jun activationdomain. Cell. 76:1025–1037.

9. Ichijo, H. 1999. From receptors to stress-activated MAP kinases. Onco-gene. 18:6087–6093.

10. Ip, Y.T., and Davis, R.J. 1998. Signal transduction by the c-Jun N-ter-minal kinase (JNK): from inflammation to development. Curr. Opin.Cell Biol. 10:205–219.

11. Lee, F.S., Hagler, J., Chen, Z.J., and Maniatis, T. 1997. Activation of theIkappaB alpha kinase complex by MEKK1, a kinase of the JNK path-way. Cell. 88:213–222.

12. Ninomiya-Tsuji, J., et al. 1999. The kinase TAK1 can activate the NIK-I kappaB as well as the MAP kinase cascade in the IL-1 signalling path-way. Nature. 398:252–256.

13. Chang, H.Y., Nishitoh, H., Yang, X., Ichijo, H., and Baltimore, D. 1998.Activation of apoptosis signal-regulating kinase 1 (ASK1) by theadapter protein Daxx. Science. 281:1860–1863.

14. Ichijo, H., et al. 1997. Induction of apoptosis by ASK1, a mammalianMAPKKK that activates SAPK/JNK and p38 signaling pathways. Sci-ence. 275:90–94.

15. Nishitoh, H., et al. 1998. ASK1 is essential for JNK/SAPK activation byTRAF2. Mol. Cell. 2:389–395.

16. Liu, H., Nishitoh, H., Ichijo, H., and Kyriakis, J.M. 2000. Activation of

apoptosis signal-regulating kinase 1 (ASK1) by tumor necrosis factorreceptor-associated factor 2 requires prior dissociation of the ASK1inhibitor thioredoxin. Mol. Cell. Biol. 20:2198–2208.

17. Saitoh, M., et al. 1998. Mammalian thioredoxin is a direct inhibitor ofapoptosis signal-regulating kinase (ASK) 1. EMBO J. 17:2596–2606.

18. Gotoh, Y., and Cooper, J.A. 1998. Reactive oxygen species- and dimer-ization-induced activation of apoptosis signal-regulating kinase 1 intumor necrosis factor-alpha signal transduction. J. Biol. Chem.273:17477–17482.

19. Zhang, L., Chen, J., and Fu, H. 1999. Suppression of apoptosis signal-regulating kinase 1-induced cell death by 14-3-3 proteins. Proc. Natl.Acad. Sci. USA. 96:8511–8515.

20. Ross, R. 1999. Atherosclerosis: an inflammatory disease. N. Engl. J. Med.340:115–126.

21. Force, T., Pombo, C.M., Avruch, J.A., Bonventre, J.V., and Kyriakis, J.M.1996. Stress-activated protein kinases in cardiovascular disease. Circ.Res. 78:947–953.

22. Lu, H.T., et al. 1999. Defective IL-12 production in mitogen-activatedprotein (MAP) kinase kinase 3 (Mkk3)-deficient mice. EMBO J.18:1845–1857.

23. Min, W., and Pober, J.S. 1997. TNF initiates E-selectin transcription inhuman endothelial cells through parallel TRAF-NF-kappa B andTRAF-RAC/CDC42-JNK-c-Jun/ATF2 pathways. J. Immunol.159:3508–3518.

24. Reed, M.A., et al. 1997. Tumor necrosis factor-a-induced E-selectinexpression is activated by the nuclear factor-kB and c-Jun N-terminalkinase/p38 mitogen-activated protein kinase pathways. J. Biol. Chem.272:2753–2761.

25. Krause, A., et al. 1998. Stress-activated protein kinase/Jun N-terminalkinase is required for interleukin (IL)-1-induced IL-6 and IL-8 geneexpression in the human epidermal carcinoma cell line KB. J. Biol.Chem. 273:23681–23689.

26. Davies, P.F. 1995. Flow-mediated endothelial mechanotransduction.Physiol. Rev. 75:519–560.

27. Traub, O., and Berk, B.C. 1998. Laminar shear stress: mechanisms bywhich endothelial cells transduce an atheroprotective force. Arterioscler.Thromb. Vasc. Biol. 18:677–685.

28. Gimbrone, M.A., Jr. 1999. Vascular endothelium, hemodynamic forces,and atherogenesis. Am. J. Pathol. 155:1–5.

29. Rothe, M., Sarma, V., Dixit, V.M., and Goeddel, D.V. 1995. TRAF2-mediated activation of NF-kappa B by TNF receptor 2 and CD40. Sci-ence. 269:1424–1427.

30. Li, Y.S., et al. 1996. The Ras-JNK pathway is involved in shear-inducedgene expression. Mol. Cell. Biol. 16:5947–5954.

31. Jo, H., et al. 1997. Differential effect of shear stress on extracellular sig-nal-regulated kinase and N-terminal Jun kinase in endothelial cells. J.Biol. Chem. 272:1395–1401.

32. Min, W., et al. 1998. The N-terminal domains target TNF receptor-associated factor-2 to the nucleus and display transcriptional regula-tory activity. J. Immunol. 161:319–324.

33. Thorson, J.A., et al. 1998. 14-3-3 proteins are required for maintenanceof Raf-1 phosphorylation and kinase activity. Mol. Cell. Biol.18:5229–5238.

34. Peterson, T., et al. 1999. Opposing effects of reactive oxygen species andcholesterol on endothelial nitric oxide synthase and endothelial cellcaveolae. Circ. Res. 85:29–37.

35. De Keulenaer, G.W., et al. 1998. Oscillatory and steady laminar shearstress differentially affect human endothelial redox-state: role of asuperoxide-producing NADH oxidase. Circ. Res. 82:1094–1101.

36. Xing, H., Zhang, S., Weinheimer, C., Kovacs, A., and Muslin, A.J. 2000.14-3-3 proteins block apoptosis and differentially regulate MAPK cas-cades. EMBO J. 19:349–358.

The Journal of Clinical Investigation | April 2001 | Volume 107 | Number 7 923